Energy management for automotive power nets

Abstract

Reducing fuel consumption has always been a major challenge to the automotive industry. Whereas first marketing aspects gave rise to innovative research, today the environmental regulations have become the main driving force behind new technologies. Historically, the research concentrated on improvements for the mechanical side of the vehicle. However, the introduction of Hybrid Electric Vehicles (HEV), where the propulsion power can also be delivered by an electric machine, definitely emphasizes the benefits of electro-mechanical solutions. With a secondary power source, the HEV can satisfy the vehicle power demand in various ways. An energy management (EM) strategy is needed to control this added freedom in a fuel-efficient way. At present, a broad range of EM strategies has been proposed in literature and several concepts have been implemented in series-production vehicles. Typically, the academic solutions focus on complex optimization techniques, arising from well defined mathematical problems. The engineering approach offers a sub-optimal strategy, based on heuristic rules. Nevertheless, both policies fail when the important vehicle characteristics for EM are not well understood. The main contribution of this thesis is to deduce a physical explanation of the EM problem for all HEV configurations, viz., the series-HEV, the parallel-HEV and the combined series/parallel-HEV. By having a good understanding of the vehicle properties of interest, it becomes possible to develop a model-based EM strategy that mimics the optimal solution, without the need for complex optimization routines, nor the necessity for having accurate predictions about the future driving cycle. The proposed causal strategy is directly suitable for on-line implementation in a vehicle. The primary goal of an EM strategy is to maximize the fuel efficiency of the vehicle. In practice, this requirement is often associated with operating the internal combustion engine (ICE) in its highest efficiency region. Nevertheless, this thesis reveals that this concept is only partially true. A better understanding of how to operate the ICE follows from two other properties: the slope of the fuel map and its fuel offset at idle speed. A formal optimization problem is formulated to prove that these properties also relate to a mathematical interpretation, and infer from the optimal solution. For all the HEV configurations mentioned above, a power-oriented vehicle model is derived. Next, a suitable EM strategy is proposed. This strategy originates from a non-causal global optimization, but through a physical understanding of the parameters of interest, it is translated into a causal on-line strategy. To cope with uncertainties in the future power demand, a feedback mechanism is added which automatically regulates the energy in the battery near a reference value. Contrary to standard control experience, this feedback control loop has a better performance if it incorporates a small bandwidth and a large tracking error. Simulation results for all HEVs demonstrate that the proposed EM strategy achieves a fuel economy which is almost equivalent to the optimal solution. Moreover, when the fuel costs for producing electric power are accurately known in advance, this strategy has the ability to further improve its performance. In practice, however, this requirement is inappropriate, since causality of the EM strategy is lost. An alternative methodology is presented to include road predictions into the causal EM strategy. By means of an electronic horizon, the prediction information is translated into a preferred reference trajectory for the energy stored in the battery. However, it will be demonstrated that the added value of having knowledge about the future driving cycle is limited, compared to the situation without prediction information. Finally, the EM concept can also be applied to the electric power net of vehicles with a traditional drive train, or micro HEVs. Here, the alternator takes the position of the electric machine. As a case-study, the EM strategy has been implemented in a Ford Mondeo vehicle. Vehicle experiments on a roller-dynamometer test-bench show that profits in fuel economy are achieved up to 2.6% for a typical driving cycle. Although the potential fuel benefits are limited for the vehicle under consideration, the return on investment is extremely high, since it requires primarily changes in the vehicle software.